Open Dots brings self-hosted agent control
Open Dots is an MIT-licensed, self-hosted workspace for chat, model connections, governed tool use, approvals, connectors, and optional computer tasks. Its README positions it as an inspectable alternative to OpenAI Dots while acknowledging its early-prototype status. [GitHub repository](https://github.com/Anil-matcha/Open-Dots)
Open Dots gets the most important agent principle right: control is a product feature, not an afterthought. Its ambition is compelling, but the project remains better suited to experimentation than production deployment. [Project documentation](https://github.com/Anil-matcha/Open-Dots)
- –Deny-by-default actions, approval prompts, and audit events give developers clearer oversight of agent behavior.
- –Configurable model endpoints and local SQLite storage support a more inspectable, self-hosted architecture.
- –Optional Docker/Playwright computer use expands capabilities while increasing security responsibilities.
- –Missing multi-user support, durable memory, scheduling, and hardened hostile-web isolation limit its replacement value today.
DISCOVERED
1h ago
2026-10-04
PUBLISHED
1h ago
2026-10-04
RELEVANCE
AUTHOR
AI Search